Most methods for jointly performing registration and super-resolution (SR) are limited to very restricted motion models, usually simple shifts. We present a method for simultaneously obtaining registration and SR from a sequence of low resolution (LR) images that can handle any global parametric motion model, also when blur is present in the LR images. We prove the validity of the algorithm with synthetic and real data experiments, being of special interest the good performance achieved in the difficult case of images registered under a projective transformation.
展开▼
